It is shown that the procedure of rejection of "erroneous" results of testing obtained in the course of evaluation of the computed mechanical and physical characteristics of soils accepted by GOST 20522-201 may lead to "unpredictable" recommendations when the requirement of increasing the maximum result of the tests leads to a decrease in the value of "computed characteristic" down to the minimum result of the tests. Some possible ways of improvement of the procedure of statistical processing of the experimental data with an aim to avoid "nonstandard" conclusions are proposed.